How RADAR rates prices

The five ratings

Every listing with enough comparable data gets a rating from 1 to 5:

• 5 · Exceptional — 30 % or more below the recent median (or 60 % or more off RRP).
• 4 · Good Deal — 15 % to 30 % below median (or 50 % to 60 % off RRP).
• 3 · Market Value — within ±15 % of the median (40 % to 50 % off RRP).
• 2 · Fair Value — 15 % to 30 % above median (20 % to 40 % off RRP).
• 1 · Above Market — more than 30 % above median (less than 20 % off RRP).

Anything more than 60 % off retail is genuinely exceptional. 50 % off is a good deal. 40 % off is market value.

How the median is calculated

1. Match the listing to a model — using brand and model aliases, fuzzy matching, and a slug-based resolver.
2. Pull the recent comparable listings for that exact model across every source we track. The window starts at 365 days. When fewer than 3 comps land inside it, we widen to 730 days before giving up — so models with sparse-but-real history (e.g. a hi-fi piece that trades a couple of times a year) still get a rating instead of a blank pill. The window we actually used is reported on the pill tooltip as Window.
3. Normalise every price to USD using daily-updated foreign exchange rates, then convert the final median to the listing's display currency.
4. Condition-adjust each comparable to an “Excellent” baseline (see below) so a Fair-condition unit doesn’t drag down the median for an Excellent listing, and vice versa.
5. Deduplicate reposts — if the same seller listed the same item three times, we count it once (averaged).
6. Take the median of the deduped sample, plus the 25th and 75th percentiles for the price range.
7. Project the median back to the listing’s own condition before computing the deal score — a “Good” listing is rated against what comparable units would sell for in “Good” condition.
8. Apply length filters for cables — lengths are matched within ±10 cm.

Condition adjustment

Condition matters. A factory-sealed amp and a well-loved one of the same model are not interchangeable, and rating them against the same median would punish honest sellers describing wear and reward those who don’t. RADAR uses a transparent multiplier table:

Condition	Multiplier
New (sealed)	1.10
Like New	1.05
Excellent (baseline)	1.00
Very Good	0.92
Good	0.82
Fair	0.65
For Parts	0.30
Unknown	1.00 (neutral)

Each comparable is divided by its own multiplier to project it to “Excellent,” the median is computed in that common space, and then multiplied by the listing’s own multiplier so the deal score reflects a like-condition comparison. When the listing’s condition is unknown, no adjustment is applied. Multipliers are conservative starting values and will be re-fitted from observed price spreads as the dataset matures.

Sold prices vs asking prices

Asking prices are aspirational. Sold prices are what the gear actually traded for. When we have 5 or more sold comparables in the rolling window, we use those exclusively — they're a much stronger signal. Below that threshold we fall back to asking prices, which are more abundant but typically run 20-40 % higher than realised sale prices.

Today, sold-price data comes mainly from StereoNET classifieds (with a SOLD flag). eBay sold-history integration is in progress, which will dramatically expand the sold corpus for AU and UK in mid-2026.

Sample size and confidence

A rating built on three listings is less reliable than one built on thirty. We handle this in two ways:

• Thin samples (2–4 unique sellers) — we blend the median rating (60 %) with an RRP-anchored rating (40 %) when retail price is known. This reduces noise on rare models.
• No median data — when fewer than 2 comparables exist but the listing states a credible RRP, we fall back to "rating vs RRP" and label the pill clearly so you know the source.
• Brand-only matches — if we can't resolve the specific model, no pill is shown. We will not rate a phono stage against a power amp.

RRP handling

Sellers can game RRP. We handle this two ways: (1) catalogue MSRP is the source of truth when available — collected from manufacturer sites, retailer listings, and authorised dealer pricing; (2) when only the seller's stated RRP is available, we boost the rating only when the discount is large enough to be unambiguous (≥40 % off). A modest seller-stated discount is treated as suggestive, not authoritative.

Currency and exchange rates

All listings are normalised to USD internally using daily mid-market rates, then converted back to the listing's display currency. This means an AUD listing in Australia can be fairly compared to a EUR listing in Germany. FX volatility introduces some noise — we expect this to be within 1-2 % on any given day.

Data sources

RADAR aggregates listings from these marketplaces:

• Audio Connection
• audio-markt.de
• eBay Australia
• eBay UK
• eBay US
• Hi Fi Exchange
• Hi-Fi Heaven
• HiFi Forsale
• HiFi NZ
• Hifitörget
• Just Audio HiFi
• Kleinanzeigen
• Len Wallis Audio
• Marktplaats
• Paragon Sight & Sound
• Reverb
• StereoNET
• The Audio Tailor
• The Music Room

We respect each source's terms of service, link back to the original listing, and don't republish photos or seller descriptions. RADAR is a price comparison engine, not a marketplace.

What RADAR does not do

• We do not authenticate products or verify sellers — buyer beware.
• We adjust pricing for stated condition but do not independently verify it — a “Good” rating on Marktplaats may not mean the same thing as on StereoNET, and we don’t inspect the goods.
• We do not consider warranty, shipping cost, import duties, or country-specific scarcity in the rating — these factors often justify paying above the global median.
• We do not rate Wanted listings or Price On Application listings.

Caveats and limitations

RADAR works best on models with deep listing history. Rare, vintage, or boutique gear may have insufficient data — we'd rather show no pill than a misleading one. Currency-converted comparisons assume goods are equally tradeable across borders, which isn't strictly true (a 100 V Japanese amplifier in a 240 V country is worth less than the median suggests).

How trending is calculated

The "What's trending" strip on the home page is a composite heat score over the last 14 days, computed per model and normalised across the candidate pool so one runaway value can't dominate. Each model gets a 0–100 score built from four orthogonal signals:

• Demand velocity — search-log hits for the model.
• Intent velocity — saved-search adds (highest weight: it's the most expensive action a user can take).
• Supply velocity — new listings appearing on the market.
• View velocity — outbound click and detail-view counts on recent listings.

Each dimension is min-max normalised across all candidate models, then combined with a weighted sum. We also overlay a price direction arrow — the change between the last two weeks of the rolling median, sourced from price_snapshots_weekly. Up ↗, down ↘, or omitted when flat.

When no signal data exists yet (e.g. fresh install), the strip falls back to a simpler "popular models this month, ranked by search count" view.

How "Best Buys" is chosen

The Best Buys strip surfaces recently-listed items priced in the cheapest quartile (market_tier = below) for their model based on the rolling weekly snapshot. Every Best Buy listing has a known image, a known condition, and a non-POA price. We over-fetch recent stock then filter — the tier itself is computed against the model's own price history, not a global threshold. This is RADAR's flagship deal-finder and has no equivalent on HiFi Shark.

Recently searched

The "Recently searched" chips on the home page show the most-recent distinct query strings logged across all users in the last 30 days, deduplicated case-insensitively, three characters or longer. We do not show who searched what — only that the search happened. This list is useful for spotting models that are getting attention right now but haven't yet built enough listing history to appear in Trending.

Found an error?

If a pill looks wrong — wrong model, wrong rating, missing data — let us know in the RADAR feedback thread on StereoNET. Every report makes the catalogue better.